[Eric Barker]

I’m no expert, but I suspect that the type of processing used to make the calculations are being done by the CPU and not the GPU, so I doubt upgrading the graphics card would have any effect. Possibly more RAM if that’s getting to be an issue (I could imagine that amount of data takes up quite a bit of space), but I think this is one of those basic CPU speed things. However, one of the really nice things about Warp Stabilizer, and it’s one of the only plugins I know that does this, is that the calculations are PERMINANT and higher on the chain to all future renders/exports. That means, unlike Rendering, which has no effect on final export speed, WS calculations are stored beforehand and don’t have to be recalculated during render or export. By the time it’s doing the actual process, it’s really performing fundamentally simple tasks (especially if it’s not set to “sunspace warp”, which I find looks weird most of the time anyway), and is very processing efficient… Basically not much more than a bunch of key frame scale, pos, and rotation animation… Possibly even lower level code than that! I’ve found it to be almost as fast at rendering than the most basic effects. Many times it yellow lines on my machine with MPE enabled.

My suggestion? Throw them on at the end of the day and let it run over night, make sure autosave is on. On my machine it takes about 30 seconds per second of a 1080i60 clip, maybe less. Unless your’re putting it on every shot of an hour-long documentary, you should be okay by morning.

Do it dead last in your editing, too. WS only calculates the frames you’ve used (including transitions outside of the in/out points). Any time you add more frames to the clip, ether by shifting the clip or increasing a transition length, you’ll need to recalculate. Sometimes it’s actually neccessary to add additional time to the beginning and end of a clip if the footage happens to be shaky during the entrance. WS assumes that the starting point is the “correct” placement, if the starting point is shaky, you may want to temporarily back up the clip to a point that is better and recalculate, then slide the in point back into place, that’ll make sure that the clip starts off very smooth and “preroll calculated”. But most of the time this isn’t neccessary.

Also, consider not using the “sunspace warp” setting or even the “perspective” setting. They tend to do some crazy things behind the scenes and CAN look kind of bizarre, depending upon the situation. I tend to use “Position, Scale, Rotation” the most.

One more thing, if you really want to go for the Nth degree of perfection , say make a steadycam shot look exactly like a jib or dolly, then consider doing some additional rotation key frame animation manually. Turn off the auto rescale property (though take note of what it was), set it manually to something larger, and get to work. Unfortunately, I’ve found that WS is not as good at rotation fixing, especially when you’re performing purposeful camera moves.

Television Producer
KTVF-11 Fairbanks, Alaska
video.ericbarker.com

[Eric Barker]

Got it! Proxy! Never had to do proxy editing before, since my work computer has always been pretty responsive and our engineers have always updated it if it starts getting slow (I should ask them to do so soon).

Before I go into detail, I just want to ask a quick question upfront: I want to find the most CPU efficient codec I can possibly have. I’m running Windows, if it makes any difference. So I’m thinking AVI 640×480 with Microsoft Video 1? Or would QuickTime JPEG/PNG be even better? Curiously, since it’s being done in the style of a silent film, the final version will be 4:3. I’m editing in 1440×1080. I ran a bunch of tests before I started this project to make sure each step along the way worked (in Premiere and After Effects), that paid off because I haven’t been faced with any gotchia’s… well, except for this!

I’m going to have to be a little careful because this video uses a bit of framerate trickery. I shot at 60p with the musicians playing the song at 2/3rds speed, then told Premiere the files were actually 90p, dragged them into a 30p sequence, and the result is 3-frame skip with an old silent-film-sped-up look that syncs with the original track. I’ve synced all the footage and it looks great, worked perfectly! Unfortunately, I may have to do some extra work with the proxy editing, since Premiere’s probably going to try and re-establish the actual framerate of the proxy files. Ahhh, I’ll re-encode them at 20fps, then load them into Premiere and tell them they’re 30fps, that way they’ll be smaller and more efficient.

This would actually be hell on normal editing, but with multicam I can get away with some things. Because of the framerate trick, I’m not going to be able to follow the normal proxy procedure of relinking the original clips. I’ll be better off Alt-dragging the clips onto their HD twins and vice-versa. I’ll need to extend the start time of every clip to its very beginning when doing this, because alt-dragging doesn’t respect the original In/Out points. But since we’re in multi-cam, that’s not much of a problem! I can just slide the entire sequence out and pull back the in points on all clips together.

Television Producer
KTVF-11 Fairbanks, Alaska
video.ericbarker.com

[Eric Barker]

I just tried it with some 60p Martial Arts footage I had laying around, and it worked great. Threw it up to 90fps, put it in a 30p sequence, and Bob’s yer uncle. Checked for any behind-the-scenes interframe blending or other BS, and none to speak of, just pulled every third frame as expected. Then I had some fun with it and threw it in an AE template that I had made that added dust, scratches, film grain, and level adjustments, not to mention a slight bit of blur (since the optics on old movie cameras wasn’t as precise as it is now), light jitter and vertical position jitter… and it looks right out of 1910… though, of course, Americans were doing Taekwondo in the teens 😉

Here’s the result!
https://www.youtube.com/watch?v=pJywpy_xr1w&feature=c4-overview&list=UU141m9J8kqdMLPo2Vj4XTpw

Television Producer
KTVF-11 Fairbanks, Alaska
video.ericbarker.com

[Eric Barker]

Good ideas, very realistic in terms of ease of putting it together. However, wouldn’t a final framerate of 30fps (progressive) just feel different from 24fps? There’s a huge difference in the feel between watching film and TV, I know that’s largely due to interlacing, and I’ve never really done a side-by-side A/B comparison, but I always figured the framerate difference just psychologically changed the flow of it. Though, doing a little more reading, it seems that some people think that is a myth, and the change in feel is more our association with film, which has many larger differences (shallow depth of field, shadow bias, non-interlacing).

However, if what you’re suggesting is a good idea, let me one-up you a bit. I think even closer would be shooting at 60fps and pulling every 3rd frame to get 20fps before transitioning to to 30fps, getting the same 1.5 ratio as 16/24. I may just be able to do all of this in Premiere by just changing the “interpret footage” framerate to 90fps, and editing on a 30p sequence! Unless it tries to do some kind of interframe blending (which it’s not supposed to unless the option is turned on), it should just pull every 3rd frame and speed it up by 150%.

Just to double-check my math, I’ll need to slow the music down to 66.67% (150% duration) before shooting in order to achieve the right final sync.

I’m going to go shoot test footage, thanks!

Television Producer
KTVF-11 Fairbanks, Alaska
video.ericbarker.com

[Eric Barker]

Interesting idea, but I’ll admit, I don’t think I’m willing to go that far, I have zero experience with real film and 10 years of daily video shooting/editing, I honestly think I’ll be able to turn out a better final movie with video. Also:

– Super 8 is probably nothing like century-old 35mm, sure they’re both celluloid, but that’s about it. Super 8 cameras run at a constant 24fps and can’t be changed, as far as I know. So while I’d be getting one step closer, I’d be taking about 3 steps back.

– I expect to do a lot of post work on this. One short section is going to require some serious rotoscoping (I’m going to make the piano look like it’s playing by itself), and I expect to do a lot of multi-take editing, I can take some time on this, but I do need to get it out in some semblence of expediency.

– My camera has a number of features which bring it close enough to film that the differences will be fairly neglagable, especially since in any case I’ll be doing some heavy treatment on it to make it look old. It has an APS-C sensor size, which is practically identical to 35mm, Super-8 is not. It has picture profiles that can bias toward shadow detail, giving a more film-like look (I actually shoot with that most of the time because its better for the harsh sunlight we get here in Interior Alaska), and as I said, I can shoot at 60p and use some method of pull down as I’m trying to decide on.

Fun idea, and I really would like to get to know film some day, but this isn’t the project. Maybe if I had access to a real hand-cranked 35mm camera. But although I do have access to a Super 8, I actually think that be because of the sensor size and frame rate, I could get closer to what I’m looking for with video.

Television Producer
KTVF-11 Fairbanks, Alaska
video.ericbarker.com

[Eric Barker]

Ack! Forgive me, it was late, I was juggling lots of figures, I totally botched that spec. 480i60 is obviously what I meant, I’ll change it in the parent.

As for “instant HD”, excuse me if I’m extremely skeptical. I’m guessing that this is high-level up conversion? We, as most production houses, have switched over to HD cameras, high resolution graphics, etc. So it’s actually much more efficient to work in HD from the beginning, except that our old sequences are still SD. I can’t imagine that even the best up conversion rivals native HD.

This conversion is more a proof of concept than anything else. It’s meant to replace our previous method of just “make a new HD sequence, copy everything over”, simply because, due to Adobe’s inability make a decent audio mixer, we’re left manually tweaking audio insert effects. However, I’d much rather spend the few minutes tweaking VSTs than rely on some up conversion software when I have perfectly good native 1080i60 elements laying around.

Television Producer
KTVF-11 Fairbanks, Alaska
video.ericbarker.com

[Eric Barker]

The audio mixer is a joke, to be sure, however, coming from audio DAWs, I’m not sure the industry standard mixer setup is going to quite make you happy, either. The one thing that video NLEs have that most DAWs don’t (though I hear Cubase has some form of this), is clip-based effects. That’s not a standard audio feature at all, and honestly, I don’t believe it should haveto be. The only reason I use clip based audio effects at all is because the mixer is so incredibly terrible. Had it been setup like any other even rudimentary DAW, there would be little need, and it would be easy to just do away with clip-based audio effects all together. A standard DAW keeps most data on the track level (not clip-level), but usually moves the data seamlessly along with the clip, so you get the best of both worlds: global track effects, but with content-aware data points.

However, Premiere’s audio mixer is beyond “dumbed down” to the point of lunacy. The fact is, so many video producers also know there way around DAWs that it would have been far more intuitive just to use the traditional DAW philosophy instead of re-inventing the wheel. now you’ve confused all of audio professionals, and made it no easier on the video-only crowd either.

The lack of true audio buses is really bizarre. Any audio engineer worth two cents sends half their tracks to aux buses for additive processing. Premiere has this, but instead of having actual bus streams, it has this bizarre “send to aux track” system, where you send the signal to a track directly, instead of a specific data stream. This may sound like a minor difference, but it makes some processing cumbersome, if not impossible. For instance, side-chaining and ducking, which are commonplace in the TV production world, are almost impossible since ducking relies on pulling a signal off of an external audio stream (bus). There is a very crazy workaround using a Russian-made VST compressor plugin, but it requires something like 6 submix tracks, and is very difficult to control.

Now, I understand that the average video editor isn’t going to be doing advanced signal processing, but a lot of the other choices Adobe has made are very unfortunate as well. One would expect that since the audio mixer uses the same audio system and effects (VSTs) as the clip-based effect system, why not just allow for loading effects directly off the effects palette? There-for you could create presets, copy data, etc. The VST architecture allows for its own preset-saving system, and somehow Adobe has chosen to completely ignore this feature.

And finally, the “only one VST open at a time” thing is beyond strange. I actually think I know what is going on, though, because I’ve been working with a VI-host programmer on the side, and know a bit about how these things are constructed. Audio effects tend to reuse the same interface window objects over and over again. Each time an effect is called up, it calls up a pre-existing window object into memory. Behind the scenes, there are actually only a few effects place-holders, and they rotate around when a new one is called up. It’s how you can switch so quickly from one sequence to another and not have to reload the audio system. However, most hosts dynamically create new effects placeholder objects on the fly every time you add one over the current limit. It’s pretty basic, really, just assign a new object and you’re done. But Premiere seems to only have one effect window object… maybe two… I think one for mono and one for stereo (these are actually completely different types of VSTs). So you open one, and it has to kill the interface object for the other to reuse the same data. Usually that’s what you do when switching sequences (that’s how MOTU does multi-sequence projects in Digital Performer, for instance) but why would you do that between different effects in the same sequence? It’s not like they take up a huge amount of processing power or RAM? After all, these same exact VSTs are being used by the dozens in a typical DAW project.

What amazes me is not that Adobe hasn’t fixed these shortcomings, it’s that they have ignored them in the face of doing a whole lot of other things that seems really inconsequential by comparison.

I do really wish the audio and video worlds would get together and learn each others’ workflows a bit more, there are aspects of each that I think would be really great to see in the other. After Effects has some KILLER workflow ideas that would be amazing in a DAW. I’d love to see javascript scripting implemented in a true DAW. Bezier curve automation is only present in a few DAWs (and even then very rudimentary). But then again, the mixing architecture of standard DAW would seem like a perfect fit in Premiere, which has completely ignored a lot of fundamental features.

Many would say, “then use a DAW for audio, and an NLE for video,” but that really shouldn’t be necessary, and is a complete cop-out. They both use the same technology, involve the same kind of interface philosophy, and it’s not as if DAWs sound better or higher quality, we’re all talking rudimentary, OS-level digital audio processing here, with the same plugin architecture. Obviously, if I want to record an album, I’m going to use a completely audio-focused environment, but it shouldn’t stop me from being able to do similar, but maybe less sophisticated, procedures an an NLE. As it is now, it feels as if Premiere is purposefully crippled to make it difficult to do standard DAW procedures. I would expect this from Avid, who wants to keep selling Pro Tools, or Apple who wants to sell Logic, but Adobe has no real foothold in the audio industry, and they’re not really trying to, so what do they have to lose?

Television Producer
KTVF-11 Fairbanks, Alaska
video.ericbarker.com

[Eric Barker]

Wow, thanks for responding! I’m guessing there’s plans of fixing this in the next update? For now I’ll just fall back to CS5 for most of my projects. Current one, I’ll re-encode all my material to something like MPEG2 iFrame. That should also make Dan Eberts happy.

Television Producer
KTVF-11 Fairbanks, Alaska
video.ericbarker.com

[Eric Barker]

First of all, number of effects isn’t going to directly influence the file size in the ways you might expect it. The largest thing to compress is full moving video. Graphics with effects is typically smaller, but these are big generalizations, it’s very complicated. What you need to know is: the file size will have NOTHING to do with how much processing power it took to create the video.

Secondly, 16MB for a 1080i file is ridiculous. If they’re truly outputting 1080i video at 16MB, the bitrate and quality must be extremely low. My guess is either, you’re comparing it to low-quality SD spots, or there’s a missing decimal point in there. 250-400MB for a 1080i file is pretty standard.

In my experience, the most common format I get from national providers is good ol’ MPEG2. I tend to see less support for H.264 in the industry, it’s a great commercial and web format, but less supported by airplay systems. And it’s an even less common transfer format. Not sure of “why”, probably mostly because it’s newer. It also doesn’t support intraframe encoding, the way MPEG2 can if necessary. At high bitrate, MPEG2s may be large, but they’re high quality and compatible with EVERYTHING, which in my experience is one of the most important things in selecting a delivery codec.

The second most common I see (but I really hate) is ProRes. We’re a Windows-based operation, and more and more houses are moving away from FCP, making ProRes a less ideal format. It works great in standard vanilla stereo mixdown, but do anything out of the ordinary, and Premiere or other non-FCP editors complete break, along with many airplay systems. I can’t tell you how much back & forth I’ve had with delivery services, trying to get the right ProRes settings that will work… it’s a great format (I wish Premiere had something of that quality out of the box) but please, keep it in house.

To be honest, Premiere’s handling of H.264 for industry delivery is downright bizarre. I’ve been playing around with it, and I can’t get it to work at all. The “MP4” family doesn’t give you any advanced settings, the “QuickTime” family is limited (though it can produce 1080i), and “H.264” while giving you all the settings, doesn’t let you create interlaced video at 1920×1080. H.264 was designed as a web and commercial media format, looks like television delivery isn’t really in it’s specs (though I don’t know why). That’s probably why I never see it.

Be prepared for large file sizes for HD video, no matter the circumstances. 10x-15x the size of SD is pretty much status quo. Our 30sec spots come out to just shy of 400MB. We render 1080i MPEG2 4:2:2 10bit at 100Mbps. This may be on the overkill end of things, but I say better safe then sorry. You should expect it, and the people you deliver to should expect it to. Lower file size = less quality. The HD world is a lot bigger and more complicated then the SD world.

I say do some research, get familiar with the different formats and their strengths/weaknesses/compatibility figure out what best suits your needs.

Also, I notice you’re shooting XDCAM, while I’m at it, I’ll give you a little tip, something that others will probably tell you not to do: there’s nothing wrong with taking the MP4 file directly out of the BAPV folder. I wrote a little script that goes through and extracts them all so I can name/organize them just like AVI files. Premiere/AfterEffects/VLC… nothing has any problem with them. I’ve been doing this for 18 months now, it’s been a godsend!

Producer Manager
KTVF-11 Fairbanks, Alaska
video.ericbarker.com

[Eric Barker]

That worked great! I modified it to randomly generate numbers instead of choosing from a list of elements, but basically, that was it.

After a little bit of studying, I understand how it works. This will help me a lot in the future when trying to figure out how to control random number generation, as well as managing time.

Thanks, Dan!

Television Producer
KTVF-11 Fairbanks, Alaska
video.ericbarker.com